QFit is a scientific analysis tool designed to streamline device characterization workflows for physicists working on superconducting quantum systems.
Responsibilities
QFit
Client
Koch Research Group at Northwestern University
Timeframe
2024
Category
Analysis tool
Problem
Device characterization is a critical but inefficient step in quantum hardware development.
Physicists must iteratively adjust model parameters to match experimental data—often bridging a ~20% gap between initial estimates and final results.
This process is:
Highly iterative and time-consuming
Fragmented across custom scripts and tools
Cognitively demanding, requiring constant comparison and adjustment
Each research group often builds its own pipeline, leading to duplicated effort & inconsistent workflows.
I Conducted user research (interviews, personas, usability testing) to identify key workflow bottlenecks and define product requirements
Project Goal
Design a tool that:
Standardizes the characterization workflow
Reduces friction in iterative parameter tuning
Supports fast comparison and decision-making
Key Insights
This is not just a visualization problem—it is a decision-making system problem.
Users are not trying to “view data,” but to:
iteratively refine parameters until model behavior matches reality.
The product must therefore:
guide users through a structured process
support rapid iteration loops
minimize cognitive overhead during comparison
Flow Chart
I mapped the end-to-end characterization process to understand the problem space. This revealed a clear pipeline, a critical Pre-fit ↔ Fit iteration loop, and opportunities to reduce friction through better structure.
Rather than fully integrating Jupyter, I designed around existing user behavior. Physicists already rely on Jupyter for model creation, so embedding it into QFit would add complexity without clear value.
Instead, I defined model creation as an external Step 0, and focused QFit on the core characterization workflow.
Solution
1. Structuring the Workflow into a Clear System
Based on the flow chart, I translated the fragmented process into a 4-step guided workflow, implemented through tab-based navigation.
Each tab represents a distinct stage in the pipeline
Provides clear orientation and progress visibility
Enables users to focus on one step at a time
This structure turns an ad-hoc process into a coherent system, making iteration more efficient and easier to manage.
2. Reducing Cognitive Load in Complex Tasks
To help users operate efficiently in a high-density environment:
Maintained layout consistency across all tabs to build familiarity
Prioritized primary actions while grouping secondary controls into collapsible panels
Improved information hierarchy to surface critical data and actions
These decisions allow users to focus on analysis and decision-making with less friction.
3. Designing for Iteration as the Core Behavior
Through research, I identified that the most critical interaction is the Pre-fit ↔ Fit iteration loop.
To support this:
Created consistent layout structures across steps to reduce re-learning
Enabled quick data transfer between stages via secondary controls
Designed interfaces that allow users to rapidly adjust, evaluate, and repeat
This transforms iteration from a fragmented process into a tight, efficient feedback loop.
Wireframing
I used wireframing to translate the workflow into a structured interface system, focusing on information hierarchy, interaction flow, and layout consistency. This resulted in a reusable layout pattern across all stages, reducing cognitive load and enabling users to iterate without re-learning the interface.
Usability Testing Insights
Usability testing revealed key gaps:
Users understood the workflow structure (4-tab model validated)
Iteration flow was effective
But:
Iconography failed to communicate domain concepts
Some controls lacked clarity and discoverability
Users needed more flexible data comparison
Iterations
I made test-driven iterations as follows:
Reorganized controls based on interaction priority (Fitts’s Law)
Replaced icons with explicit terminology + status indicators
Enabled multi-graph comparison for complex use cases
Clarified ambiguous actions through visual direction cues
These changes improved clarity, efficiency, and trust in the system.
AI-Assisted Visual Exploration
After stabilizing the core workflow through multiple rounds of UX iteration, I shifted focus to refining the visual system.
To accelerate exploration, I translated research insights and usability findings into structured prompts, using AI tools to rapidly explore and prototype multiple design directions before I analyzed the pros and cons for each style.
Compact Grid Style
Pros: Clean, structured, and visually organized
Cons: Inefficient use of space in parameter controls; weak support for frequent graph interactions
Terminal Style
Pros: Familiar to advanced users; low visual noise and high efficiency
Cons: Poor discoverability; less accessible for users with mixed experience levels
Dashboard Style
Pros: Strong parameter grouping through card-based layout; improved clarity and operational confidence
Cons: Introduced non-essential features; reduced efficiency in comparing parameter values
Final Design
Through this process, I synthesized the most effective elements into a final design that prioritizes clarity, efficiency, and alignment with user workflows.
Outcome
Reduced workflow time by ~90% in user testing
Standardized a previously fragmented process into a clear, repeatable system
Open-sourced tool adopted within the research community
QFit demonstrates how UX design can transform a complex technical workflow into a structured, efficient decision-making system, thus recognized with Core77 Design Award 2024 — Notable Apps and Platforms
QFit is currently available online and is open-sourced under BSD-3-clause license. This software is not for profit, because the development of the app is supported by governmental research grants, and part of the code is powered by Python library scQubits, which is also open-source under BSD-3-clause license.
